Patients and tissue samples
A cohort of 176 unselected primary OCCC patients was enrolled at the Department of Pathology, Fudan University Shanghai Cancer Center from 1999 to 2016. Tumor specimens were collected after obtaining written consent from all patients with the approval of the Facility Ethical Committee (Fudan University Cancer Center Ethics Committee, China; approval No. 050432-4-121B, 13 December 2012).
All cases were reviewed, and the diagnosis was confirmed by two pathologists (Dr Ge and Dr Bi). All 176 cases were pure OCCCs, including two cases of synchronous endometrial endometrioid carcinoma. The following clinical data were extracted from the medical history and examination records: age, tumor size, and Federation of Gynecology and Obstetrics (FIGO) stage. All international FIGO stages were reclassified according to the 2014 FIGO guidelines.
Tissue microarray (TMA) construction
Hematoxylin-eosin (HE)-stained slides were reviewed by Dr Ge and Dr Bi, and an abundant tumor cell area was chosen for TMA construction.
The TMA was prepared using triplicate core samples from 176 OCCCs to achieve a high level of standardization for immunohistochemical analysis. Triplicate 1-mm tissue cores were taken from each targeted donor block, and a receptor block was inserted. Four-micrometer sections were cut and stained with H&E to confirm the histologic diagnosis. Unstained slides were used for immunohistochemical staining.
Immunohistochemical staining was performed on TMAs with the Ventana Benchmark XT platform. The following panel of antibodies was used: anti-MLH1 (clone G168-728, Roche), -PMS2 (clone EPR3947, Roche), -MSH2 (clone G219-1129, Roche), -MSH6 (clone 44, Roche), and -ARID1A (clone D2A8U, Roche). According to the published criteria of abnormal expression of MMR and ARID1A , both markers were interpreted as loss of expression, whereas no tumor cells were stained.
We chose the recommended guidelines that the immunohistochemistry results be reported in a binary manner, either positive (indicating intact mismatch repair, showing intact nuclear expression in tumor cells) or negative (indicating deficient mismatch repair, showing nuclear expression completely lost in tumor cells) [ Nuclei of lymphocytes and ovarian stromal cells served as positive internal controls. All staining results were reviewed by the above 2 pathologists. Because immunohistochemical examination was performed using a tissue microarray, when deficient mismatch repair (dMMR) was observed, the immunohistochemistry procedure was repeated on whole-slide sections to avoid heterogeneous expression, such as for MSH6. Ultimately, MMR protein restaining was performed in 51 instances involving 27 of 176 cases; ARID1A restaining was performed in 15 instances involving 3 cases of dMMR among 135 cases.
All tumor slides were evaluated by Dr Ge and Dr Bi, who were blinded to the MMR status, and an average of 7 tumor slides per case (range 1-17) were examined. The following features were assessed (in brief).
Host inflammatory response
(1) TILs – defined only as lymphocytes located within the boundary of tumor cells or glands – were counted, especially in regions of high tumor density and minimal stroma. Areas showing the greatest number of lymphocytes at low-power magnification were randomly selected, and the numbers of lymphocytes were counted in 10 high-power microscopic fields (HPFs). Apoptotic bodies and stromal lymphocytes were not counted (Fig 1A).
(2) Intratumoral stromal inflammation – defined as inflammation in the stroma between tumor cells, nests, or glands that is evident at low-power magnification – was graded as diffuse (marked inflammatory infiltrate present in multiple foci or on multiple slides), focal (partially present in the stroma with inflammation) or absent (8). (Fig 1B). Areas of necrosis and infarction were excluded.
(3) Peritumoral lymphocytes – defined as peritumoral lymphocytes present at low-power magnification – were categorized as present or absent (Fig 1C).
(4) Plasma cell infiltration – defined as the percentage of plasma cells in the intratumoral stroma – was classified as < 10%, 10% to 50%, or > 50% (Fig 1D).
Figure 1 Tumor-infiltrating lymphocytes (TILs) (A, arrows) in tumor cells, diffuse intratumoral stromal inflammation (B), composed mostly of plasma cells (C), and peritumoral lymphocytes (D) were observed.
(1) Background precursor – evaluated based mainly on the benign element adjacent to the tumor in the same section, categorized as a tumor arising as endometriosis, adenofibroma, both, or neither.
(2) Architectural pattern – included tubulocystic, papillary, glandular, solid or mixed.
(3) Stromal hyalinization – graded as diffuse, focal, or absent.
(4) Nuclear atypia – high cytologic grade demonstrating variations of greater than 3 times the nuclear size, highly irregular nuclear contours, striking hyperchromasia, and/or prominent nucleoli. Areas constituting at least 10% of the tumor were considered high grade; other areas were considered low grade.
(5) Specific morphology (signet ring cells) – defined as present or absent.
(6) Tumor necrosis – defined as present when there were geographic foci of tumor cell necrosis. Areas of necrosis only within glands or at the tumor's surface were not included.
Overall survival (OS) was defined as the time from the operation to either death or the last follow-up. Disease-free survival (DFS) was defined as the interval from the operation to disease recurrence or the last follow-up. We defined disease recurrence as a consistent elevation in CA125 or detection of a tumor by a clinical examination, including a physical examination and imaging. The patients were followed up until December 2018 by medical records or telephone.
Statistical analysis was conducted using SPSS version 20.0 software (IBM, SPSS Statistics Armonk, NY, and USA). Univariate analysis of categorical variables was performed with the chi-square test and Fisher’s exact test; multivariate analysis was performed with exact logistic regression to identify independent predictors of abnormal MMR expression. Survival analysis was performed using the Kaplan-Meier univariate method. Differences in survival curves were determined with log-rank tests. P values < 0.05 were considered statistically significant.